Embracing AI’s Evolution: The Rise of Personalized Software in Scholarly Publishing

AI is evolving at an astonishing pace. In just the past few weeks, we’ve seen major advances in agent-based software development, new model releases, and a growing push toward AI-driven automation. These developments point to a future where software is no longer just built for users but shaped around them—what GitHub’s CEO recently described as the age of personalized software.

This shift carries profound implications for scholarly publishing, particularly for the large, monolithic platforms that currently dominate the space. If this trajectory holds, we could see a radical transformation in the next two years—one where AI moves beyond simply assisting software development to generating fully functional applications from documentation and user intent.

In fact, much of this is already possible. I’ve been experimenting over the last months with workflows that translate development requirements into software at remarkable speed. Instead of long development cycles, we can rapidly iterate and deliver working applications, with engineers focusing on architecture, testing, scalability and quality. This methodology prioritizes delivering value first, with product quality evolving in subsequent iterations. The result is a highly agile approach that allows us to build, iterate, and deploy quickly—ensuring continuous improvement while keeping up with the accelerating pace of AI-driven development.

As AI’s reliability in test-driven development improves, the need for manual oversight will likely diminish further, paving the way for even more dynamic and responsive publishing systems. I can see a day coming soon where simply creating a bug report in an environment like GitHub will trigger an AI agent that not only identifies and fixes the issue but deploys the solution to a staging environment for testing—completing entire development cycles in minutes rather than days or weeks.

Impacts on Journals and Publishing Platforms

For journals, publishers, and infrastructure providers, this shift presents both opportunities and challenges. Today’s dominant publishing platforms are designed to accommodate a wide range of use cases, often resulting in large, complex systems that struggle to adapt quickly to emerging needs. But as AI-driven software creation becomes the norm, we may see the rise of lighter, more adaptable solutions—custom-built for specific journal workflows, editorial needs, and publishing models.

Imagine a future where publishers can rely on AI to develop submission and review platforms tailored precisely to their editorial processes, or where AI augments existing systems with seamless integrations and refinements. Instead of forcing journals into rigid, pre-built platforms, AI could enable a more modular approach—where editorial workflows and peer review processes are dynamically configured based on journal-specific needs.

This would allow journals and publishers to move away from reliance on all-in-one solutions and toward a more flexible publishing ecosystem, where infrastructure is designed and refined in real-time. It could mean a shift from platforms designed for broad user bases to publishing tools that are continuously optimized to support evolving workflows, new business models, and emerging research dissemination practices.

Challenges and Considerations

Despite the promise of AI-driven personalization, transitioning from entrenched, monolithic platforms won’t be easy. Large publishing systems, with their extensive infrastructure and deeply embedded workflows, will face inertia in adapting. Moreover, cultural resistance—both from developers and users—could slow the shift toward AI-driven software generation.

Yet the potential upside is undeniable. AI’s ability to interpret product documentation, infer needs, and rapidly generate applications could make publishing platforms more intuitive, flexible, and responsive. Instead of forcing journals and publishers to conform to legacy systems, AI-powered solutions could evolve dynamically, aligning with the changing landscape of scholarly publishing.

That’s why embracing an iterative, value-first methodology is crucial. We can already leverage AI to create functional publishing platforms quickly, then refine and optimize them based on real-world feedback. This approach reduces the risk associated with major platform overhauls while enabling innovation at a pace that traditional development cycles simply cannot match.

The Road Ahead: Expert-Driven Development (ED)

This shift is leading us into an era of what I’ve previously described as Expert-Driven Development (EDD)—a methodology where the experts in the use case are the developers. In this model, AI removes the barriers between domain experts and software creation, allowing those with deep knowledge of publishing workflows—editors, journal managers, and production staff—to directly shape the tools they use. Instead of relying on intermediaries to translate requirements into software, they can interact with AI-driven systems that generate, test, and refine publishing infrastructure based on their needs.

This represents a fundamental change in how scholarly publishing platforms are developed. The emphasis moves away from lengthy development cycles dictated by traditional engineering teams and toward an iterative, expert-led process where publishing professionals, engineering teams and AI work together to literally build and refine the systems they rely on. The game of empathy boards is over. The developers are the end users via AI and engineers are the folks that make sure it scales.

We are on the verge of a transformation where publishing infrastructure will no longer be dictated by large, static platforms but will evolve in real-time, shaped by those who know the workflows best. This is not a distant vision—it’s happening now, and the next few years will reveal just how powerful this approach can be.